What we work on
Risk-based pricing for consumer lenders
Risk-based pricing is a marketing model that sets the price a customer is offered from that customer’s risk. Your own risk scores are the input; it builds none. Take a consumer lender: four rates in place of one lift a month’s first-year margin from €126,500 to €202,900, on 212 fewer loans.
Try it
Set a rate for each band and watch margin move
The rates stay as set. Today’s margin stays put, because nothing changes at today’s rate.
Margin against today
+€76,400
€202,900 against €126,500, first year, on a month's loans
Loans against today
−212
818 against 1,030 loans a month
Loss against today
−€40,900
€167,350 against €208,250, first year, on a month's loans
Ahead of today until take-up reacts about 2.8 times as strongly as assumed, then behind it. At this response, no band earns less than it does today.
Margin by band, today against the rates set
The same data as a table
| Band | Rate | Take-up | Loans | Yearly loss | Margin today | Margin at this rate | Change |
|---|---|---|---|---|---|---|---|
| A | 11% | 50% | 300 | 0.96% | €66,000 | €68,100 | +€2,100 |
| B | 14% | 38% | 266 | 2.7% | €70,000 | €77,140 | +€7,140 |
| C | 19% | 32% | 160 | 7.04% | €15,000 | €51,680 | +€36,680 |
| D | 20% | 46% | 92 | 13.2% | −€24,500 | €5,980 | +€30,480 |
| All | 818 | €126,500 | €202,900 | +€76,400 |
The full demo adds the chart of where your rates fall behind today as take-up reacts more strongly.
What it is
What risk-based pricing is
A single rate charges every approved applicant the same price for different risks. Risk-based pricing sets a rate for each risk band, so each band’s price carries that band’s loss.
The work starts after the credit decision. Your risk scores, your bands, the loss you expect in each and the decision to approve or decline all stay with your risk team. The model starts from those numbers and answers one marketing question: what to offer the applicants you have already approved.
The question
What to offer each risk band
For each band, the analysis sets a rate and shows what that rate does to three numbers:
- Take-up: the share of approved applicants who accept at that rate.
- Margin: what the band earns after funding, servicing and loss.
- Loss: what the band loses, because a higher rate changes who accepts as well as how many.
Take-up by rate is the number to measure. It comes from past rate changes, or from a rate test that your compliance team has cleared. A test shows two rates inside one band, so the warning on customers who compare prices applies to it.
How many applicants each rate needs is a two-group comparison, the one the price-test calculator sizes.
Example
A worked example
A consumer lender approves 2,000 applicants a month and offers all of them 12%. A loan carries €5,000 over the year; funding and servicing cost 5.5% of that. Its own scores sort applicants into four bands, and its own policy caps any rate at 20%.
| Band | Approved a month | Yearly loss | Take-up at 12% | Take-up lost per point of rate | Margin left at 12% |
|---|---|---|---|---|---|
| A | 600 | 1% | 40% | 10 points | 5.5% |
| B | 700 | 2.5% | 50% | 6 points | 4% |
| C | 500 | 5.5% | 60% | 4 points | 1% |
| D | 200 | 10% | 70% | 3 points | −3.5% |
The last column is the rate less cost and loss. Band D loses money at 12%, and band A pays well above its risk. The model adds one assumption of its own: a band’s loss rises by 4% of itself for each point on the rate, because the applicants who accept a dearer loan are not a random slice of the band.
Trying every rate in half-point steps up to the cap gives each band the rate with the most margin. Band A pays less than today, the other three pay more, and band D stops at the cap.
Band A pays less than today, the other three more, and every rate clears what its band costs
The same data as a table
| Band | Rate | Funding, servicing and loss |
|---|---|---|
| A | 11% | 6.46% |
| B | 14% | 8.2% |
| C | 19% | 12.54% |
| D | 20% | 18.7% |
| Band | Rate | Take-up | Loans | Yearly loss | Margin today | Margin at this rate |
|---|---|---|---|---|---|---|
| A | 11% | 50% | 300 | 0.96% | €66,000 | €68,100 |
| B | 14% | 38% | 266 | 2.7% | €70,000 | €77,140 |
| C | 19% | 32% | 160 | 7.04% | €15,000 | €51,680 |
| D | 20% | 46% | 92 | 13.2% | −€24,500 | €5,980 |
| All | 818 | €126,500 | €202,900 |
Loans
1,030 to 818
loans a month, today against the band rates
Margin
+€76,400
€126,500 to €202,900, first year, on a month's loans
Loss
−€40,900
€208,250 to €167,350, first year, on a month's loans
Total loss falls because fewer loans are written, mostly in the riskiest bands. Loss per loan does not fall: it moves from €202 to €205.
A higher average rate explains part of that, so the comparison needs a third line: the best single rate, 13.5%. Moving everyone there earns €141,088. About 81% of the gain comes from setting rates by band, and the rest from the higher average rate.
Margin under the three schemes: about 81% of the gain comes from setting rates by band
The same data as a table
| Scheme | Margin | Against today |
|---|---|---|
| Today: 12% for all | €126,500 | €0 |
| Best single rate: 13.5% | €141,088 | +€14,588 |
| A rate for each band | €202,900 | +€76,400 |
| Scheme | Loans | Loss | Margin |
|---|---|---|---|
| Today: 12% for all | 1,030 | €208,250 | €126,500 |
| Best single rate: 13.5% | 838 | €194,113 | €141,088 |
| A rate for each band | 818 | €167,350 | €202,900 |
| The same rates, take-up responding half as strongly | 924 | €212,060 | €232,440 |
| The same rates, take-up responding twice as strongly | 606 | €77,930 | €143,820 |
The band rates fall behind today at about 2.8 times the assumed response
We would recommend these rates only on measured take-up. They stay ahead of today with take-up responding twice as strongly, and they fall behind it at about 2.8 times. The lender also gives up 212 loans a month, which is the wrong trade if volume is the target. The example holds the balance level for the year, treats take-up as a straight line in the rate, and leaves competitors out; a real read-out tests each of those.
The rules
Two rules on risk-based pricing and credit
As read on 4 November 2026. These are facts about the rules, for your counsel to apply.
United States. Risk-based pricing based on a consumer report carries a notice rule (15 USC 1681m(h), 12 CFR 1022.72). The rule does not restrict the pricing, and a credit-score disclosure to all applicants is the alternative to the notice (12 CFR 1022.74).
European Union. Directive 2023/2225 applies from 20 November 2026. It has no risk-based pricing notice. Article 13 and Recital 46 require telling consumers of a personalised offer or price from automated processing. On the creditworthiness assessment, which stays with your risk team, Article 18(3) keeps social networks and special-category data out of it and Article 18(8) gives human review where it is automated. Article 31 covers excessive rates. We have not checked how each country has transposed it.
Whether a gap between two customers’ rates is defensible, beyond these notices, is in When is charging two customers different prices defensible? Where an AI agent sets or recommends the rate, test it against its instruction: Who does your AI agent work for?
The work
Who it is for, and what you receive
Done for you: the agents and the analytics director work as your data team, from FinDaS’s side, and nothing is installed on your systems.
Who it is for. Consumer lenders that already score their applicants and have no data team on the pricing side.
What you need to have. Your risk score or band for each approved applicant, the loss your risk team expects in each band, loss by rate where past rate changes show it, the offers made and accepted with rate and amount, and your funding and servicing cost.
The method, in plain words. Take-up by band and rate first, from past rate changes or a cleared test. Then margin and loss at each rate, against today and against the best single rate. Then the rate for each band, within your ceiling, with the point where it stops paying.
What you receive. A rate memo of up to two pages: the rate for each band, what it does to take-up, margin and loss, and where it stops holding. A method note listing the checks that ran. If a test is needed, a test-design memo before it.
How it is checked. Agents build the analysis and draft the memos. Every deliverable gets number and claim checks run by scripts, not AI, and a review by an agent. On what you act on, that review is a panel of three reviewer agents and a judge, and then the analytics director signs it. The checks are described on how it works.
Your data. Where it goes, which model providers process it and what happens when the work ends: see data handling.
Price. Fixed per deliverable. A rate review by band, designed and read out, is a sprint, from €12k, fixed in a one-page proposal before work starts. Repeat work runs on an async retainer, from €8k a month.
Record. The analytics director has built and run risk-based pricing before, as marketing work at a consumer-finance company: a model that set the price each client was offered from that client’s risk. That work did not include the credit-risk models.
Not covered. Credit scoring, credit-risk models, underwriting policy and legal advice. Price tests and personalised offers are on pricing analysis.
Start here
Setting rates by risk band?
Up to thirty minutes on the question and your data, then a fixed proposal.